Talk to your data!
VerbaGPT aims to make data analytics using large language models easy without compromising data privacy.
You can ask questions of your CSV or SQL data in natural language and get answers quickly. See VerbaGPT in action below:
See more examples using menu at the top-right of this page.
Try a Live Demo of VerbaGPT
More Than Talk…
More answers, more complex questions. Go beyond simple data aggregation queries. Get the ability to make plots, ask complex queries, and even perform data modeling. See example videos of VerbaGPT in action (menu, top-right).
No Sharing Data…
Your data is more important than ever. Don’t give away your strategic asset by giving an Ai vendor or an LLM access to your data.
VerbaGPT can run locally on your hardware. VerbaGPT’s distinctive architecture guarantees that the Large Language Model (LLM) is never granted direct access to your data. Instead, only the database schema—specifically, table and column names—are disclosed to the LLM, along with any column-related details that you, the user, expressly authorize for sharing.
On the product roadmap is incorporating local models (like LLama), which would allow for completely offline analytics.
Simple and Intuitive…
VerbaGPT presents an easy way to harness the power of LLMs for tabular data exploration. We have designed it so that the focus is on you getting the most from your data, and pushing all the complexity of generative Ai to the background.
Enough talk, see it in action!
See videos here. We will be posting more video examples of VerbaGPT in action regularly. If you are interested in requesting access to VerbaGPT, or having us do a demo for you, or just want to share your thoughts, come right in.
-
Complex Classification
In a previous post, we asked VerbaGPT to build a classification model for us. In this example we use VerbaGPT to classify the type of Iris flower from the famous iris dataset. However, we challenge it further by asking for not one but 3 models, and giving us a confusion matrix as the measure of…
-
Classification: Iris
In this example we use VerbaGPT to classify the type of Iris flower from the famous iris dataset. Video: Coming soon… User question: Give me a model to classify the iris flower. Give me the confusion matrix as a performance metric. VerbaGPT response: Cost of query: $0.00 [[50 0 0] [ 0 47 3] […
-
Climate Data
In this example we use VerbaGPT to analyze some climate data. We use publicly available data from IMF. Video: Coming soon… We start by asking a simple data question. User question: Give me the temperature change in USA for 2015. VerbaGPT response: Cost of query: $0.00 Temperature change in USA for 2015: 1.531 Nice, clean…
-
Multiple Models
In this example we ask VerbaGPT to give us the top 5 models to predict a variable. Video: User question: Give me the top 5 models to predict Rings. use r2 as performance metric VerbaGPT response: Cost of query: $0.00 Random Forest: 0.5727 Gradient Boosting: 0.5616 Linear Regression: 0.5390 XGBoost: 0.5213 Decision Tree: 0.1661 This…
-
Rank Top Drivers
In this example we ask VerbaGPT to give us a ranked list of variables that drive a target variable’s value. In this example we use the famous (in data modeling) Abalone dataset . Video: User question: Rank the top 5 variables that explain the variation in Rings VerbaGPT response: Cost of query: $0.00 Feature F-value…
-
SQL Joins + Plot
In this example we ask VerbaGPT a plotting question that requires a non-trivial SQL query. Video of VerbaGPT in action: User question: Give me a bar plot of sales, by album, of U2 VerbaGPT response: Cost of query: $0.00 Answer: VerbaGPT identified the correct database (“Chinook”), identified relevant tables, performed the necessary joins and answered…
-
Vague/Complex Ask
In this example we ask a vague question that isn’t straightforward, and requires joining tables in a relational database (Chinook) to give a response. Let’s see how VerbaGPT does. When I asked the question, I had “U2” in mind. Video: User question: Rank the top selling albums by that famous irish band. VerbaGPT response: Cost…
-
SQL Table Joins
In this example we ask a a question that requires joining tables in a relational database (Chinook) to give a response. Video: Coming soon… User question: Give me the top 10 selling artists. VerbaGPT response: Cost of query: $0.00 Answer: (‘Iron Maiden’, 140) (‘U2’, 107) (‘Metallica’, 91) (‘Led Zeppelin’, 87) (‘Os Paralamas Do Sucesso’, 45)…
-
Linear Regression
In this example we ask VerbaGPT to give us a model and associated performance metrics on some data. Video: User question: Give me a model to predict Rings. Provide at least two performance metrics, including r2 VerbaGPT response: Cost of query: $0.00 Answer: Mean squared error: 5.01 R-squared: 0.54 VerbaGPT answered the question correctly. Since…
-
Basic Plot
In this example we ask VerbaGPT to produce a simple plot. Video: Coming soon… User question: Give me a histogram for Rings. VerbaGPT response: Cost of query: $0.00 Answer: VerbaGPT answered the question correctly. Since VerbaGPT comes with an embedding framework, it did not have to be told which database (i.e., Db_Test_01*) to look into,…
Interested? Talk to us!